Erratum to The Legacy of Political Violence across Generations

Abstract

In the original publication of Lupu and Peisakhin (2017), we miscoded one of the dependent variables in our analysis. Our measure of turnout is a factored index of two items from our survey asking about participation in the March 2014 referendum and the September 2014 local elections. Our survey instrument coded these two variables differently so that the turnout index is actually reversed, with higher values corresponding to individuals who were less likely to participate. This error was an oversight on our part—we incorrectly presumed that our instrument had used the same value labels for yes/no responses.

The result of this error is that the effect of turnout throughout the article is inverted. The magnitude and statistical significance of the effect remains unchanged. Corrected versions of Figures 2 and 5 from the original article can be found here:

As a result of this correction, we find mixed results regarding the effect of ancestor victimization on political engagement: while victimization reduced turnout in the two elections we examined, it increased respondents’ willingness to participate.

The article's main claim is that ancestor victimization strengthens ingroup attachment and animosity toward the perpetrator within families that experienced more state repression. We demonstrated how the mechanism behind this effect is the transmission of victim identities across multiple generations. The set of findings at the core of the original article is unaffected.

In measuring how victim identities affect political participation, one of the variables we examined was turnout in two 2014 elections. The other relevant variable was willingness to participate in other political activities, like protests and petitions. We found that ancestor victimization increases willingness to protest (this result is unchanged). Owing to the coding error, we reported that descendants of victims are more likely to turn out to vote when they are, in fact, less likely to do so.

In 2014, Crimean Tatar leaders urged their community to boycott the Russia-backed elections that followed the region's annexation. It makes sense that those with stronger group attachments (the descendants of more intensely victimized families) would have been more likely to heed the call for a boycott, and therefore, less likely to turn out, and we presented our incorrect positive result as somewhat surprising. As a result, the revised finding on political participation is in some ways more consistent with our core argument. At the same time, given their animosity toward Russian authorities, it also makes sense that the descendants of victims would be more willing to participate in protests and petitions in the future.

We have revised the supporting information and replication dataset to correct this error. We are grateful to Austin Wang for bringing it to our attention.